As 2025 comes to an end, many teams working in computational biology are noticing a shift. Projects are getting larger, more multi-modal and longer lived. Results matter, but so do continuity, reproducibility and the ability to evolve analyses over time as data, methods and teams change.
Much of this work happened quietly inside projects, omics workflows and build systems. Taken together, it reflects a broader move away from one-off analyses towards systems that scale, remain understandable and can be relied upon over the long term.
Atlas building & multi-modal work
Several projects moved decisively into atlas-style work, combining spatial, single-cell and other modalities into shared, evolving resources. For many organisations, the challenge here is no longer generating an initial result, but maintaining coherence as datasets grow and in-house resources need to be dynamically updated over months or years.
Supporting this kind of work requires workflows that are modular, versioned and designed to be revisited and extended. Instead of treating each analysis and modality as a fresh start, teams increasingly need structures that preserve context and intent over time.
What emerged this year is a clearer pattern for building and maintaining atlas-like resources that remain usable well beyond their first release. This shift is becoming essential as multi-modal data moves from exploration into sustained discovery.
From pipeline to platform: high-throughput RNA-seq automation
High-throughput RNA-seq remains a backbone of many research programs, but scaling it reliably is often harder than expected. One notable outcome this year was seeing our high-throughput RNA-seq workflow has matured into a flagship, production-grade system.
What sets it apart is not a single feature, but how it is designed to be used in practice: fully automated from raw data to results, while supporting evolving protocols and experimental design. Instead of being tuned for one study, the workflow is set up to be extended, maintained and reused across teams.
For organisations running RNA-seq at scale, this kind of maturity turns a pipeline into dependable infrastructure - something that can be trusted to keep working as datasets grow and requirements change.
Viash Hub as the new default
As computational projects grow, the bottleneck is not method development, but knowing what analyses exist, how to run them, and whether they will keep working over time. This year, Viash Hub became the default way we address all three.
First, it provides a clear, searchable overview of available analyses, so teams can see what already exists before building something new. Second, it makes those analyses runnable in a consistent way, with clear instructions that reduce setup time and eliminate environment-specific guesswork.
Most importantly, Viash Hub provides the infrastructure for continuous integration and automated builds. This is what turns collections of scripts into dependable systems: every change is built, tested, and released in a controlled way, ensuring that workflows remain reproducible and maintainable as interfaces evolve and dependencies change.
Taken together, this shifts computational work from ad hoc execution to a sustainable default: discover, run, and maintain workflows with confidence, even as teams, data, and requirements change.
Contributing to the ecosystem
Alongside client delivery, continued investment in open source and upstream projects remained an important part of the year. Contribution to efforts such as OpenProblems, LaminR, AnndataR and even fixes in Nextflow itself reflect a commitment to working at the level of where workflows are shaped, not just consumed.
This kind of deep technical work rarely draws attention, but it strengthens the foundations that many teams rely on. It also helps ensure that practical lessons from real-world projects feed back into the tools and standards used across the community.
Looking ahead
Looking ahead, the focus will shift from proving that these systems work to making them easier to use, extend and adopt. A large part of that effort will go into improving the usability of Viash Hub - reducing friction in how workflows are discovered, configured and shared, and making the underlying infrastructure more accessible to a broader range of users.
At the same time, more attention will move toward integrating multi-modal datasets in a systematic way. As projects increasingly span modalities, the challenge is no longer just running analyses side by side, but connecting them through shared structure, metadata and versioned workflows that support long-term evolution.
The aim for the coming year is to continue strengthening the foundations: turning mature workflows into clear products, and making complex, multi-modal analysis feel less bespoke and more like a dependable, repeatable process.